Broadly speaking, artificial intelligence encompasses many different approaches, the main idea of which is to make a program act like an intelligent being to solve problems. Machine learning is an approach to AI that does not rely solely on pre-design, but rather summarizes from data to achieve a simulation of memory and reasoning. It includes such as Support Vector Machines (SVM), various types of algorithms based on decision trees (including Boosting, Bagging, Random Forest, etc.), various types of algorithms based on artificial neural networks (e.g., Simple Networks and Deep Networks, etc.), as well as the integration of multiple methods.

Based on the advantages of the development of artificial intelligence, many partners want to make great achievements in this field, but the three thresholds in front of you need to be overcome one by one. In this article, Qianfeng shares with you the three barriers to get started in artificial intelligence.

Threshold one, the foundation of mathematics

We should have understood, whether for big data or for artificial intelligence, in fact, the core is the data, through the organization of the data, analysis of data to achieve, so the math has become a mandatory course for the introduction of artificial intelligence!

Mathematical technical knowledge can be divided into three major disciplines to learn:

1, linear algebra, very important, modeling calculations rely on it ~ must review solid, if not normally used may forget more;

2, high + probability, these two as long as you master the basics on the line, such as integration and derivation, a variety of distributions, parameter estimation and so on.

The importance of probability and mathematical statistics is mentioned, because almost all algorithms in cs229 are deduced from parameter estimation and its significance in the probability model, and the updating rules of the parameters are probabilistically interpretable. For work on the design and improvement of algorithms, generalization is a core course, bar none. When given off-the-shelf algorithms, only a basic knowledge of probability is needed to read and understand them, and then a more extensive knowledge of line generations is needed to make the model run efficiently.

3, statistics-related fundamentals

Regression analysis (linear regression, L1/L2 canonical, PCA/LDA dimensionality reduction)

Cluster analysis (K-Means)

Distribution (normal distribution, t-distribution, density function)

Indicators (covariance, ROC curve, AUC, coefficient of variation, F1- Score)

Significance test (t-test, z-test, chi-square test)

A/B test

Threshold II, English proficiency

The English I am talking about here is not the English four or six, we all know that computers originate from abroad, and a lot of valuable literature comes from abroad, so if you want to make achievements in the direction of AI. Still need to read some foreign literature, so to achieve a level of English that can read and understand foreign literature.

Threshold three, programming technology

First of all, as a general programmer, C++ / Java / Python such a language skill stack should be essential, of which Python needs to focus on the application of crawlers, numerical computation, data visualization.

The three thresholds of AI entry are some essential basics, so don't mind the hassle, it's crucial to build a good foundation!